[PATCH 2/2] [RFC] Add checkpoint/restart support for epoll files.

Matt Helsley matthltc at us.ibm.com
Tue Aug 25 05:09:47 PDT 2009


On Tue, Aug 25, 2009 at 12:17:08AM -0400, Oren Laadan wrote:
> 
> 
> Matt Helsley wrote:
> > On Mon, Aug 24, 2009 at 04:27:25PM -0500, Serge E. Hallyn wrote:
> >> Quoting Matt Helsley (matthltc at us.ibm.com):
> >>> Save/restore epoll items during checkpoint/restart respectively.
> 
> [...]
> 
> >>> +/* TODO confirm that get_current_user() has been restored */
> >> Not sure what you mean.  At this point, the task's credentials
> >> are still the ones used when it called sys_restart().  We won't
> >> update them until the end of the restart operation.  However the
> >> normal file restore operations should have reset the file->f_cred
> >> to the checkpointed value - including hierarchical user namespaces.
> > 
> > I wanted to check to ensure that we're properly enforcing the epoll
> > watch limits for the restarted tasks. So the patch looks good because
> > it restore the "common" file pieces before the epoll items/watches.
> > See below if you're curious where/how that happens. I think I can
> > remove this TODO now.
> 
> Good point. Me thinks it deserves a comment in the code ?

Easy enough. :)

Cheers,
	-Matt Helsley


More information about the Containers mailing list